Subir Roy: Long and short of semiconductors

VALUE FOR MONEY

The long attempt to bring semiconductor manufacturing to India (currently India is non-existent in this field) has received a fillip with a successful visit by a delegation of the Taiwan Semiconductor Association. A couple of deals are expected to be announced within six months and in the words of one of the eager delegation members, "we should have come five years ago".

Since Taiwan is a global leader in semiconductor manufacturing, this is good news for India. Indian expertise in software is globally established, so if it can tie up with the most sophisticated of hardware guys from Taiwan, then it will be an Asian marriage made in heaven. That may well happen one day but a long period of getting to know each other in a distinctly western way has to come first.

Let's look at the good news first. Semiconductor manufacturing can be broken up into three parts: designing, assembly and testing, and, the most daunting, wafer manufacturing, termed fabrication, undertaken by "fabs", which need a billion dollars and more to set up. Says the president of the Taiwan Semiconductor Industry Association, T Y Wu, "In India the fab is zero. Some packaging and testing houses have started but technologically at the low end. In design Indian companies have a lot of strength, in certain areas even better than Taiwan because of software skills. So design is clearly the one where companies can find a match, find things on which to collaborate for future benefits. That will evolve more than the packaging and testing houses."

The not so good news is that India remains as far away as it has been for some time in getting a fab of its own. A South Korean company signed an MoU with the Andhra Pradesh government for the purpose when Chandrababu Naidu was chief minister and was allotted 50 acres. There has been no progress on that project and, according to reports, the second-hand plant that Intel was to sell to the Koreans for setting up in India has been sold to the Chinese.

The other project is the one mooted by SemIndia. It also has been coming for a long time. Touted as a $3 billion complex housed in an area named FabCity. The promoters of FabCity asked for 1,200 acres and were eventually allocated 290 acres by the Andhra government. AMD, a global semiconductor leader, has announced a technical collaboration with SemIndia, but, despite reports to the contrary, AMD's involvement with the project till now remains at the technological level, without any commitment of funds. The promoters of SemIndia, on the other hand, do not seem to have much funding of their own. The cause of semiconductor manufacturing in India would have been served better if its forward scout had more financial clout than SemIndia.

But there is a strong, even compelling, case to locate fabs in India. Semiconductor devices, or chips as they are popularly called, sit at the heart of electronics manufactures, account for a significant part of the final value of such products, and also are among the most complex items to go into them. To aspire to become a global IT power and not have fabs is like trying to run a race on one leg. You can, but you will clearly remain an also ran.

With India emerging as one of the key global markets of tomorrow with a huge appetite for a new generation of affordable electronic devices, it makes sense to get into the fab business. Fabs facilitate the setting up of hosts of supporting industries, boost applications development, and generally help grow the knowledge base. Also, if you don't get into fabs soon enough, you will be left out of a new generation of manufacturing technology. Fabs churning out wafers of 90-65 nanometer specification are coming and if you are not there you are out of the cutting edge of activity.

It would be lovely to have fabs but they cost upwards of a billion dollars and have to be retooled every three or four years, needing change of around 90 per cent of the machinery. The business of chips is a high-volume, low- margin one and the global semiconductor market is very volatile. Samsung, the aggressive Korean technology leader and investor, lost dearly at the turn of the last decade over semiconductors. This is the reason why most countries which are into this high-tech game have subsidised the setting up of fabs. Recently, Israel has subsidised Intel's fab of $1.6 billion with a help of $608 million. Germany has footed ¤545 million for AMD's plant in Dresden in former East Germany. China subsidises a host of its new industries, including fabs. The one country which does not subsidise fabs any more is Taiwan, whose semiconductor industry has matured.

The Indian Semiconductor Association has therefore urged the government of India to set aside $1.5 billion, to be disbursed over the next three years, to kickstart the industry. Plus, a long period of tax holiday has been proposed. There are two main arguments against subsidising fabs today. One, trade barriers are low in the age of globalisation when no country thinks of becoming entirely self-sufficient. With the Indian market developing the way it is, sooner or later global firms will come forward to set up fabs in India out of their business compulsion. Till then just allow zero duty imports.

More importantly, since this is an industry where technological obsolescence comes fast, only a large player can be expected to come in and keep investing for upgrade over time. The way in which Intel goes round the world shopping for subsidies, it could set up a plant anywhere. If location is unimportant to it with its proprietary technology, how much spin off benefit will it bring to the rest of industry just by coming? Israel and Germany are industrially developed enough to made quick and best use of a fab. What good did Japan do to Malaysia's eco-system by dismantling Japanese fabs and bringing them to Malaysia years ago? Should India wait with its subsidy package a little longer until its testing and assembly units mature and large Indian players with deep pockets come forward to set up fabs? The debate is not over yet.